Optimal Primary Tone Levels in Distortion Product Otoacoustic Emissions and the Role of Middle Ear Transmission. Steven Charles Marcrum

Transcription

1 Optimal Primary Tone Levels in Distortion Product Otoacoustic Emissions and the Role of Middle Ear Transmission Steven Charles Marcrum

2

3

4

5 Acknowledgments This dissertation is the culmination of years-long efforts on the part of numerous individuals, to all of whom I owe a great deal. I am particularly indebted to Prof. Dr. Werner Hemmert, without whom this dissertation would not have been possible. His openness and generosity were more than I could have hoped for. I am grateful to Prof. Dr. Peter Kummer, whose steadfast vision for the potential and future of otoacoustic emissions inspired this work. Furthermore, I thank him for the countless hours and incredible energy he invested in my development not only as a researcher, but also as a colleague. A heartfelt thanks is due to Dr. Thomas Steffens for the many productive conversations had in his office, on trains, in restaurants, and most everywhere else I found him. His rigor and uncompromising dedication to the scientific method has shaped this dissertation. Additionally, I would like to thank Christoph Kreitmayer for his data collection efforts and Dr. Florian Kandzia, Dr. Andre Lodwig, and Dr. Thomas Rosner for the technical assistance they so kindly provided. Finally, I would like to thank Christina Wallner, whose faith, energy, and patience always seemed to pick up where mine left off. Regensburg, April 2017 v

6 vi

7 The truth is not distorted here, but rather a distortion is used to get at truth. Flannery O Connor vii

8 viii

9 Contents List of Abbreviations xii Abstract xvi 1 Introduction DPOAE generation Primary tone optimization Outline Optimal DPOAE Primary Tone Levels in Normal Hearing Adults Introduction Methods Participants Equipment & stimuli Evaluation of probe signal stability Experiment 1: Evaluation of L DP reliability Experiment 2: Development of an optimization formula Data analysis Results Evaluation of probe signal stability ix

10 Contents Experiment 1: Evaluation of L DP reliability Experiment 2: Development of an optimization formula Importance of primary tone level optimization Discussion Acoustic Immittance Measures in the Prediction of Optimal DPOAE Primary Tone Levels Introduction Hz tympanometry Wideband energy absorbance Methods Participants Equipment & calibration Procedures Multivariable model of optimal primary tone levels Data analyses Results Tympanometry Energy absorbance Multivariable model creation and performance Discussion Limitations Conclusions Estimation of Minor Conductive Hearing Loss Using Distortion Product Otoacoustic Emissions Introduction x

11 Contents 4.2 Methods Participants Equipment & calibration DPOAE stimuli Procedures Data analyses Results Effect of CHL on optimization formula parameters Accuracy of CHL DP estimates Discussion Limitations Conclusions Conclusions 98 Bibliography 103 List of Figures 121 List of Tables 124 xi

12 List of Abbreviations ABG ABR ANOVA BM cc CHL CHL DP CHL P T dapa db DPOAE EA Expt air-bone gap auditory brainstem response analysis of variance basilar membrane cubic centimeter conductive hearing loss DPOAE-based estimate of CHL pure tone audiometry-based estimate of CHL dekapascal decibel distortion product otoacoustic emission energy absorbance experiment xii

13 F f 1 f 2 female frequency of the lower-frequency primary tone frequency of the higher-frequency primary tone f 2param value of the multivariable model parameter associated with f 2 f DP FFT GM HL I/O khz L 1OP T L L 1 frequency of the DPOAE fast-fourier transform geometric mean hearing level input/output kilohertz change in L 1 needed to recover an optimal L 1 -L 2 relationship left level of the lower-frequency primary tone L 1ERROR error in the prediction of L 1OP T L 1OP T optimal L 1 L 2 L DP L DP M AX M level of the higher-frequency primary tone level of the distortion product maximal L DP observed within a given series male xiii

14 LIST OF ABBREVIATIONS ml mm n NA NH OAE OHC PC pespl R RM ANOVA SD sec SEM SNR SPL TPP TW UHF milliliter millimeter number not applicable normal hearing otoacoustic emission outer hair cell personal computer peak equivalent sound pressure level right repeated measures analysis of variance standard deviation second standard error of measurement signal-to-noise ratio sound pressure level tympanometric peak pressure tympanometric width ulta-high frequency xiv

15 V ea Y tm equivalent ear canal volume peak-compensated static acoustic admittance xv

16 Abstract Distortion product otoacoustic emissions (DPOAEs) are low-level, audio frequency signals emitted from the cochlea in response to sound, which are measurable within the external ear canal. DPOAEs represent an objective means of assessing the integrity of active cochlear mechanics and are evoked in response to stimulation with two sinusoids, or primary tones, of levels L 1 and L 2. Despite great progress towards predicting the primary tone level relationships which maximize DPOAE amplitude, and thereby clinical utility, in the average ear, significant inaccuracies are routinely observed when attempting to predict optimal characteristics within any given ear. Individual differences in middle ear energy transmission, capable of affecting both the absolute, as well as relative, primary tone level relationships effective within the cochlea, represent an as yet unaccounted for contributor to the predictive difficulties. The purpose of this dissertation was to evaluate the relationship between ear-specific middle ear energy transmission characteristics and optimal DPOAE stimulation parameters with an eye towards increasing personalization, and thereby accuracy, of predicted optimal stimulus levels, expanding the clinical utility of DPOAEs in ears both with normal hearing and minor conductive hearing loss, and increasing insight into the basic mechanisms of DPOAE function in humans. To that end, both univariate and multivariable primary tone level optimization formulas were developed from a sample of 30 participants (57 ears) xvi

17 with normal hearing. In the univariate model [L 1 = 0.49L (db SPL)], the average optimal L 1 is predicted for each L 2 in the traditional manner, irrespective of the potential characteristics of a given middle ear. In the multivariable model [L 1 = 0.47L EA + f 2param + 38 (db SPL)], the L 1 recommendation is influenced not only by L 2, but also by an ear- and frequency-specific measure of energy absorbance into the middle ear and the primary tone frequency. Results suggest that use of the multivariable formula leads to statistically significant reductions in L 1 recommendation error, as compared to the univariate formula. In contrast to the mean improvement (mean = 0.18 db, SD = 1.54 db), which was too small to be considered clinically meaningful, sizable improvements in L 1 recommendation accuracy were identified within individual ears. Though generally weak in the absence of measurable conductive hearing loss (CHL), a stronger relationship between middle ear function and optimal primary tone levels was identified in the presence of mild CHL (3-10 db). For a single ear of 30 adults with normal hearing, the effect on auditory threshold of increased air pressure within the ear canal was estimated via comparisons between optimal DPOAE primary tone level relationships determined both in the presence and absence of the excess air pressure. A highly significant linear dependence was identified between DPOAE- and pure tone audiometry-based estimates of CHL, r(19) = 0.71, p < However, the correlation was only significant when ear-specific optimization formula parameters were known. Viewed together, the preceding studies suggest that, for ears presenting normal middle ear function, differences in middle ear energy transmission, as quantified using clinical measures, do not meaningfully influence optimal DPOAE primary tone level relationships on average. However, significant effects can occur in individual ears. Mild conductive hearing loss, on the other hand, has a significant impact on optimal separations. Further, this effect can be exploited xvii

18 under certain conditions, constituting an additional objective source of information regarding middle ear health. xviii

19 1 Introduction Otoacoustic emissions (OAEs) are low-level, audio frequency signals emitted from the cochlea in response to sound, which are measureable within the external ear canal. First recorded in 1978 (Kemp, 1978), OAEs have since revolutionized the understanding of cochlear function, though their origins, as well as clinical potential, are still not fully appreciated. OAEs convey a wealth of information regarding the integrity of active inner ear and middle ear mechanics and have, consequently, found their largest impact in the area of clinical hearing assessment. Distortion product otoacoustic emissions (DPOAEs), a tonal subtype of OAE resulting from intermodulation distortion produced by nonlinear aspects of cochlear processing, are evoked through the simultaneous presentation of two pure tones and have proven useful for myriad clinical purposes. Specifically, DPOAEs have shown utility as an objective means of identifying normal and hearing impaired ears, such as for use in newborn hearing screening programs or in the evaluation of other difficult-to-test populations (Gorga et al., 1997; Musiek and Baran, 1997; Johnson et al., 2007, 2010; Kirby et al., 2011), estimating hearing threshold (Boege and Janssen, 2002; Gorga et al., 2003; Oswald and Janssen, 2003; Janssen et al., 2005), differentiating sensorineural and conductive hearing loss (CHL) (Gehr et al., 2004; Janssen et al., 2005; Janssen, 2013), differentiating cochlear and neural pathology for the diagnosis of auditory neuropathy/synaptopathy (Hood, 2015), 1

20 Introduction and monitoring for deleterious side-effects of ototoxic medications (Reavis et al., 2011; Konrad-Martin et al., 2012), among other uses. Recently, DPOAEs were also shown to potentially be of use for the objective quantification of conductive hearing loss (Kummer et al., 2006; Olzowy et al., 2010; Deppe et al., 2013). It is this latter role, predicated upon the existence of a systematic relationship between CHL magnitude and DPOAE amplitude, which is of primary interest for this dissertation. In particular, the relationship between ear-specific middle ear energy transmission characteristics and optimal DPOAE stimulation parameters was assessed with an eye towards increasing personalization, and thereby accuracy, of predicted optimal stimulus levels, expanding the clinical utility of DPOAEs in ears both with normal hearing and minor conductive hearing loss, and increasing insight into the basic mechanisms of DPOAE function in humans. Figure 1.1 presents a schematic representation of a typical computer-based DPOAE measurement system, which includes a probe for stimulus presentation and response collection, digital signal processing (DSP) unit for stimulus generation and response processing, and personal computer (PC) for measurement control. The probe, which contains two miniature receivers and a low-noise microphone, is sealed within the ear canal using a soft rubber ear tip. The DSP unit synthesizes two pure tone signals of frequencies f 1 and f 2, which are subsequently routed to separate receivers within the probe following digital to analog conversion. Independent receivers are used in an effort to prevent the creation of artificial intermodulation distortion products, which can potentially be observed when driving a single receiver by both the f 1 and f 2 signals simultaneously. In this way, the signals are first mixed acoustically within the ear canal, as opposed to electrically at some earlier stage in the stimulation process. The cochlear response, as well as any biological, system, or external noise present within the ear canal, is 2

21 collected by the microphone and fed to the DSP unit following analog to digital conversion. Within the DSP unit, a number of signal processing techniques can be applied in an attempt to optimize detection of the DPOAE, with signal averaging being prevalent. Additionally, fast-fourier transformation is utilized to allow for comparison of the level of the signal+noise in the frequency bin corresponding to the desired distortion product with those surrounding bins containing only noise. A valid DPOAE result is obtained when the level of the signal+noise exceeds the average level of the surrounding noise by a criterion amount or when some other set of measurement quality control standards are met. 3

22 Introduction Figure 1.1: Schematic representation of a typical computer-based DPOAE measurement system. Sinusoids of frequencies f 1 and f 2 are synthesized and presented via miniature receivers sealed within an ear level probe. A low-noise microphone collects the response and feeds it to a digital signal processing unit for noise reduction and further processing. Reprinted from Diagnosis of hearing disorders and screening using artificial neural networks based on distortion product otoacoustic emissions. In Lim, C. T. & Hong, J. C. H (eds.), 13th International Conference on Biomedical Engineering, Jyothiraj, V. P. & Kumar, A. S., 2009, pp Copyright 2009, Springer Verlag - Berlin Heidelberg. 4

23 1.1 DPOAE generation Graphical output of the DPOAE system to the user following measurement within a normal hearing ear is presented in Figure 1.2. The filled circles represent sound pressure levels within 2f 1 -f 2 frequency bins in response to stimuli with frequencies f 1 and f 2 (f 2 /f 1 = 1.22) and levels L 1 = 65 and L 2 = 55 (db SPL). The location of each circle along the abscissa is determined by the f 2 of the associated stimulus pair. Though dependent upon numerous technical, as well as physiological, factors, average DPOAE levels in response to moderate level stimuli range between approximately 5 25 db SPL and are generally db below the levels of the stimulus tones. The open circles represent the average level of the 5 frequency bins on either side of the bin containing 2f 1 -f 2, or the noise present at the time of measurement. Due to the spectral characteristics of ambient and physiological sounds, noise levels are frequently observed to rise as DPOAE frequency decreases. If the obtained DPOAE amplitudes were purely quantifications of the DPOAE signal, any observed DPOAE level could be taken as evidence of a cochlear response. However, these values reflect the combined contributions of both the DPOAE, as well as any noise present within the given 2f 1 -f 2 bin. Clinical convention dictates that an arithmetic difference between DPOAE and noise levels of approximately 6 db is sufficient to acceptably limit the contribution of any noise to the overall DPOAE response level, though the arbitrary nature of this rule should be noted. Depending on the aims of the particular study, differences of 12 db or more are often preferred for research purposes. 1.1 DPOAE generation A simplified, one-dimensional transmission-line model, which aids in understanding the bi-directional acoustical / mechanical signal flow between the ear canal 5

24 Introduction Figure 1.2: Diagnostic DPOAE results obtained within a normal hearing ear. Filled circles represent the levels within the various 2f 1 -f 2 frequency bins. Open circles represent the average level of the 5 frequency bins on either side of the bin containing 2f 1 -f 2, or the noise. and cochlea during the DPOAE generation and measurement process, is presented in Figure 1.3. Sound pressure (P e ), as observed at any location within the ear canal, is obtained through summation of all pressure waves propagating forward towards the tympanic membrane with those propagating away from the tympanic membrane (P e = P + e +P e ), with superscript + and - denoting forward and reverse transmission, respectively. Of note, these waves consist not only of the direct-path signals from the receivers and DPOAE generation region, but also any pressure waves arriving at the measurement location following reflection. The si- 6

25 1.1 DPOAE generation nusoids used to evoke DPOAEs, known as primary tones, are calibrated for level within the ear canal and are of frequencies f 1 and f 2 (f 1 <f 2 ) and levels L 1 and L 2 (L 1 L 2 ). The acoustic pressures of the presented tones act on the tympanic membrane (P o = P + o +P o ), resulting in mechanical vibrations propagating through the middle ear, which serves as an impedance matching system. Specifically, the effective areal difference between the tympanic membrane and the stapes footplate and the lever constituted by the length of the manubrium of the malleus relative to that of the long process of the incus result, when combined with the middle ear s frequency-dependent spring-mass properties, in a frequency-specific pressure gain, which approximates that loss attributable to the higher impedance of the cochlear fluids as compared to the air within the ear canal (Aibara et al., 2001). This approximately 29-fold pressure increase applied at the base of the cochlea (P b = P + b +P b ) allows for the efficient creation of hydromechanical waves within the fluid of the inner ear. These waves subsequently establish a pressure differential across the basilar membrane (P c = P + c +P c ), setting up traveling waves along its surface which peak at the characteristic place of each primary tone frequency. It is currently hypothesized that DPOAEs, as measured within the ear canal, are the vector sum of the products of two distinct generation mechanisms, which are depicted schematically in Figure 1.4. The primary mechanism is an intermodulation distortion mediated by nonlinear aspects of outer hair cell (OHC) transduction, which occurs at the location of maximal overlap between the two traveling waves, or the f 2 region. Consensus regarding the specific nature of the distortive mechanism has not yet been achieved, though a joint effect of OHC (Brownell, 1990) and stereocillia (Liberman et al., 2004) electromotility is suspected. The secondary generator is known as the coherent-reflection mechanism and can be understood as impedance perturbances on the basilar membrane in the area of the 7

26 Introduction Figure 1.3: Schematic representation of signal flow through the ear canal, middle ear, and cochlea. Superscript + and - represent forward and reverse signal transmission, respectively. Adapted from Theory of forward and reverse middle-ear transmission applied to otoacoustic emissions in infant and adult ears, D. Keefe & C. Abdala, 2007, J Acoust Soc Am, 121(2), p Copyright 2007, Acoustical Society of America. tonotopic place of each distortion product (Shaffer et al., 2003; Shera, 2004). As a given distortion product is created, its associated pressure wave spreads basally along the basilar membrane from the f 2 region towards the ear canal (P c ), as well as apically towards the DPOAE s tonotopic place (P + a ), where it is partially reflected and re-directed towards the ear canal (P a ). Therefore, the DPOAE observed at the probe microphone constitutes a mixture of direct and reflected signals, which have driven the middle ear system in reverse and become measurable at the distortion product frequency. The phase of the product of the distortion mechanism has been shown relatively invariant with frequency, while the component resulting from the coherent-reflection mechanism exhibits a steep phase gradient (Talmadge et al., 1998, 1999; Mauermann et al., 1999; Shera and Guinan, 1999). Due to this difference between the components in terms of the rate of phase change as a function of frequency, quasi-sinusoidal patterns of constructive 8

27 1.1 DPOAE generation Figure 1.4: Schematic representation of the nonlinear distortion and coherentreflection mechanisms of DPOAE generation. Compressive nonlinearities in the overlap region of the two stimulus traveling waves, f 1 and f 2, generate distortion products which spread along the basilar membrane in both directions. In this example, energy of the 2f 1 -f 2 distortion product travels basally from the overlap region towards the ear canal, as well as apically towards its tonotopic place, where it is partially reflected by impedance discontinuities. Reprinted from Sources and Mechanisms of DPOAE Generation: Implications for the Prediction of Auditory Sensitivity, L. Shaffer et al., 2003, Ear Hear, 24, p Copyright 2003, Lippincott Williams. and destructive interference can emerge. The level of the observed DPOAE, and thereby the clinical utility of the measure itself, depends greatly upon the phase relationship of the two components at the frequency of interest, with significant peaks or dips, known as fine structure, occurring for frequencies at which interference approaches its extremes. Recent research efforts on the topic of fine structure have largely focused on developing methods to limit the effects of interference on DPOAE amplitude, whether through segregation of the distortion and reflection components in the time domain (Vetesnik et al., 2009; Dalhoff et al., 2013; Zelle et al., 2013) or through use of an additional tone to suppress the reflection component (Heitmann et al., 1998; Talmadge et al., 1999; Dhar and Shaffer, 2004; 9

28 Introduction Johnson et al., 2007; Kirby et al., 2011). However, multiple studies have also assessed the potential clinical utility of analyzing characteristics of the overall fine structure itself (Brown et al., 1993; Engdahl and Kemp, 1996; Rao and Long, 2011; McMillan et al., 2012; Poling et al., 2014). For example, Engdahl and Kemp (1996) reported significantly reduced fine structure depth in human ears following exposure to moderate-level noise. DPOAE fine structure might therefore prove a sensitive means of detecting noise-induced hearing damage, prior to its impacting subjective auditory thresholds. Despite both areas of research having produced promising results, much is yet to be done, as there currently exists neither a widelyaccepted clinical method for reducing fine structure nor a deep understanding of the implications of its presence. The largest and most commonly utilized DPOAE in humans occurs at the cubic difference frequency 2f 1 -f 2, though multiple mathematically-related distortion products are frequently observable in response to stimulation with a single pair of primary tones (2f 1 -f 2, 3f 1-2f 2, 4f 1-3f 2, etc., as well as 2f 2 -f 1, 3f 2-2f 1, 4f 2-3f 1, etc.). Figure 1.5 displays the various distortion products measurable within one particular normal hearing ear in response to stimulation with primary tones of frequencies f 1 = khz and f 2 = khz. In addition to the clinically-utilized distortion product at 2f 1 -f 2 (dashed line), other signals are observable at 3f 1-2f 2 and 2f 2 -f 1 (dotted lines), though the potential clinical utility of these additional distortion products is not well-understood. Attempts to include distortion products at frequencies other than 2f 1 -f 2 in clinical protocols, such as for the objective identification of hearing status, have met with limited success (Gorga et al., 2000; Fitzgerald and Prieve, 2005; Kirby et al., 2011). 10

29 1.1 DPOAE generation Figure 1.5: Results of a DPOAE measurement with f 1 = khz and f 2 = khz (gray lines). Numerous distortion products are created simultaneously, including the clinically-meaningful DPOAE at 2f 1 -f 2 (dashed line). Though other distortion products are also apparent (dotted lines), they are currently of limited clinical utility. 11

30 Introduction 1.2 Primary tone optimization The basilar membrane (BM) can be conceptualized as a frequency analyzer, in that traveling waves resulting from signals of differing frequency achieve their maximum displacements at specific and roughly distinct locations along its surface (Plomp, 1964). This frequency resolving capability results partly from the membrane s passive mechanical properties. Specifically, the basilar membrane becomes progressively wider and less stiff from base to apex, with these gradients resulting in the base of the BM responding with the largest displacements to high-frequency signals, while the apex responds to low-frequency signals best. However, this mechanism in isolation is not capable of producing the high-degree of frequency resolution associated with normal basilar membrane function. Rather, an active mechanism, commonly described as the cochlear amplifier (Davis, 1983), is additionally needed. The term cochlear amplifier refers to the nonlinear, frequency-selective amplification of the traveling wave by means of outer hair cell electromotility and stereociliary active bundle movements. In the case of DPOAEs, distortion generation is maximal when BM displacement due to stimulation with the two primary tones is equivalent near the characteristic place of f 2. However, the frequency-selective nature of the cochlear amplifier dictates that growth of the basilar membrane response with alteration of stimulus level will differ between the BM locations maximally responding to the two primary tones. Specifically, growth for the f 2 primary tone will be significantly more compressive than for the f 1 primary tone, which generally peaks between 0.25 and.50 octaves below f 2. Figure 1.6 shows how the growth of basilar membrane displacement velocity changes as a function of distance from the measurement location (characteristic place for 8.5 khz). For test frequencies surrounding the 12

31 1.2 Primary tone optimization measurement location, growth is compressive. However, for frequencies below 7 khz or above approximately 11 khz, growth becomes more linear. Equalizing displacement resulting from the primary tones, and thereby optimizing primary tone characteristics, therefore necessitates an increasing level difference between the primary tones as the level for f 2 is decreased. Numerous studies have attempted to identify the DPOAE stimulus parameter relationships which optimize traveling wave overlap, and therefore DPOAE level, for the average ear. Table 1.1 displays the results of a systematic review summarizing available studies having investigated the effect of primary tone frequency and level relationships on the level of the 2f 1 -f 2 distortion product. While results in humans have generally revealed an optimal primary tone frequency ratio of approximately f 2 /f 1 = (Harris et al., 1989), findings in terms of optimal level separations of the primary tones have been more divergent (Whitehead et al., 1995; Stover et al., 1996; Kummer et al., 1998; Neely et al., 2005; Johnson et al., 2006). However, despite variability in the specific parameter values suggested, more recent primary tone level optimization formulas tend to be consistent in their recommendations of increasing L 1 -L 2 separation with decreasing L 2. Kummer et al. (1998), for example, suggested the optimization formula L 1 = 0.4L [db SPL], thereby recommending a 0.4 db reduction in L 1, and therefore a 0.6 db increase in L 1 -L 2, for each 1 db reduction in L 2. However, several trends are also apparent in the available datasets, which serve to threaten their utility and necessitate further research. First, sample sizes have tended to be small, with 14 of 32 studies consisting of 10 or fewer participants. This limitation serves not only to restrict the generalizability of findings, but also the range of statistical methods available for use during data analysis. For the set of identified studies, inferential statistics were provided in only 10 of 32 reports. Second, perhaps due to 13

32 Introduction the significant time investment involved in more complete experimental protocols, the majority of studies investigated only a narrow range of the available L 1,L 2 space and f 2, leaving open the possibility of different behavior in other parameter regions. Third, in spite of the recognized significance of primary tone level separation for the optimal generation of DPOAEs and the potential for middle ear characteristics to impact this separation through alteration of primary tone levels during forward transmission (see Figure 1.3), no primary tone level optimization formula currently attempts to account for ear- and frequency-specific middle ear effects. Aural acoustic immittance measures, such as 226-Hz tympanometry and wideband energy absorbance (EA), represent time-efficient methods through which the signal transmission properties of specific ears can be quantified. Indeed, both have been successfully implemented clinically for the differentiation of healthy ears and those exhibiting middle ear pathology, such as otitis media with effusion (Marchant et al., 1986; Johansen et al., 2000; Beers et al., 2010). Of significance for the present work, owing to relatively high test-restest reliability, these measures can even be used for the assessment of differential transmission characteristics within healthy middle ears. Incorporating tests of middle ear energy transmission into the development of primary tone level optimization formulas could therefore allow for a more accurate estimate of the L 1 -L 2 effective within a given, as opposed to average, cochlea, and thereby simultaneously contribute not only towards a better understanding of middle ear function, but also allow for an ear-specific customization of recommended DPOAE primary tone levels. Furthermore, improved understanding of the nature of the systematic relationship between middle ear and primary tone characteristics could potentially be leveraged for diagnostic use, such as the objective identification of CHL and estimation of its magnitude. 14

33 1.2 Primary tone optimization Figure 1.6: Basilar membrane velocity input-output functions obtained for several frequencies at the tonotopic place for 8.5 khz in chinchilla. The dashed line represents linearity. Functions for frequencies around 8.5 khz display more compression than those below 7 khz or above 11 khz. Primary tone pairs will be increasingly affected by this difference in growth rate as f 2 -f 1 increases. Adapted from Basilar membrane mechanics at the base of the chinchilla cochlea. I. Input-output functions, tuning curves, and response phases, Robles et al., 1986, J Acoust Soc Am, 80(5), p Copyright 1986, Acoustical Society of America. 15

34 Table 1.1: Studies investigating the influence of primary tone frequency and level relationships on the level of the 2f 1 -f 2 DPOAE. ANOVA analysis of variance, db decibel, Expt experiment, F female, GM geometric mean, HL hearing level, Kruskal-Wallis H Kruskal-Wallis one-way analysis of variance, L left, M male, n number, NA not applicable, NH normal hearing, R right, RM ANOVA repeated measures analysis of variance, SPL sound pressure level, UHF ultra-high frequency. Reprinted from A systematic review of stimulus parameters for eliciting distortion product otoacoustic emissions from adult humans, L. Peterson et al., 2017, Int J Audiol, doi: / Copyright 2017, Taylor & Francis.

35

36

37

38 Introduction 1.3 Outline Optimal DPOAE primary tone levels in normal-hearing adults: In this chapter, theory underlying optimal stimulation of distortion product OAEs is introduced and published primary tone level optimization formulas are reviewed. In Experiment 1, relevant technical and physiological sources of variability, heretofore not acknowledged in DPOAE optimization formula development, are evaluated. Specifically, a clinical DPOAE system is assessed for signal stability, reliability, and stimulation accuracy. Additionally, test-retest reliability of evoked DPOAE levels is assessed in an effort to quantify, among other features, the reliability of the physiological mechanisms of DPOAE generation. In Experiment 2, frequency-specific and nonspecific DPOAE primary tone level optimization formulas are developed, which incorporate the findings of Experiment 1. Finally, formula performance is compared with that of formulas currently utilized clinically (see Chapter 2). Acoustic immittance measures in the prediction of optimal DPOAE primary tone levels: In this chapter, 226-Hz tympanometry and wideband energy absorbance measures are evaluated in terms of their utility for the improvement of primary tone level recommendation accuracy, resulting in the inclusion of energy absorbance results into a multivariable model. Recommendation accuracy is then compared between the multivariable formula developed here and the univariate formula developed in Chapter 2. Additionally, normative ranges for 226-Hz tympanometry and wideband energy absorbance are presented. (see Chapter 3). Estimation of minor conductive hearing loss in humans using distortion product otoacoustic emissions: In this chapter, the feasibility of objectively quantifying experimentally-produced, minor CHL in humans is assessed by comparing CHL estimates resulting from DPOAE- and traditional pure tone 20

39 1.3 Outline audiometry-based methods. Additionally, the accuracy of DPOAE-based CHL estimates obtained when using generic, as opposed to ear-specific, optimal primary tone level formula parameters is investigated. The method s potential for clinical implementation is discussed (see Chapter 4). Conclusions: The primary findings of this dissertation are reviewed and discussed in context. An effort is made to highlight potential clinical implications and directions for future research (see Chapter 5). 21

40 2 Optimal DPOAE Primary Tone Levels in Normal Hearing Adults A version of the following chapter first appeared as the peer-reviewed article Average optimal DPOAE primary tone levels in normal-hearing adults, S.C. Marcrum et al., 2016, Int J Audiol, 55, p Copyright 2016, Taylor & Francis. 2.1 Introduction Distortion product otoacoustic emissions (DPOAEs) are low-level signals, which are emitted from the cochlea in response to the simultaneous presentation of two primary tones. Primary tones are sinusoids presented at frequencies f 1 and f 2 (f 2 > f 1 ) and levels L 1 and L 2 (L 1 L 2 ). Current understanding of DPOAE generation suggests that they originate from two sources. The primary source is a compressive nonlinearity in basilar membrane (BM) mechanics near the f 2 characteristic place. The second is the coherent-reflection mechanism, which reflects the apically-spreading energy of the 2f 1 -f 2 distortion product at its tonotopic place (Shaffer et al., 2003). Due to differences between the products of these sources in terms of the rate of phase change as a function of frequency, quasi-sinusoidal patterns of constructive and destructive interference can emerge. The level of the 22

41 2.1 Introduction DPOAE as measured within a given ear can depend greatly upon the phase relationship of the two products at the frequency of interest. Clinically-significant peaks or dips, known as fine structure, can occur for frequencies at which interference approaches its extremes. Across many ears, however, systematic effects of phase will be reduced and DPOAE level will be greatest when overlap of the BM excitation patterns for the primary tones is maximized near the f 2 place (Shaffer et al., 2003; Young et al., 2012). As excitation patterns along the BM are strongly affected by the functional state of outer hair cells (OHC) and OHC function is reduced in hearing loss, DPOAEs have been found useful as an objective means of identifying hearing impairment (Gorga et al., 1993; Kim et al., 1996; Stover et al., 1996; Gorga et al., 1997; Musiek and Baran, 1997; Dorn et al., 1999; Johnson et al., 2007, 2010; Kirby et al., 2011), estimating subjective auditory threshold (Nelson and Kimberley, 1992; Suckfull et al., 1996; Boege and Janssen, 2002; Gorga et al., 2003; Oswald and Janssen, 2003; Janssen et al., 2005; Johnson et al., 2010), and even quantifying minor conductive hearing losses (Kummer et al., 2006; Olzowy et al., 2010). Optimizing primary tone characteristics is essential for maximizing the level, and therefore utility, of DPOAEs. Gaskill and Brown (1990) reported that for a given f 2 /f 1 ratio and L 2, the level of the DPOAE (L DP ) will be maximized in response to a certain optimal L 1 (L 1OP T ) and will be reduced in response to other values of L 1. Multiple investigations have since reported optimization formulas attempting to predict L 1OP T in both normal and hearing impaired ears (Whitehead et al., 1995; Stover et al., 1996; Kummer et al., 1998; Neely et al., 2005; Johnson et al., 2006). Utilizing data from Gaskill and Brown (1990), Kummer et al. (1998) identified the relationship L 1 = 0.4L [db SPL] as that which maximizes DPOAE level. In an extensive follow-up investigation with the f 2 /f 1 ratio fixed 23

42 OPTIMAL DPOAE PRIMARY TONE LEVELS at 1.2 in accordance with the findings of Harris et al. (1989), Kummer et al. (2000) similarly reported L 1 = 0.4L [db SPL] as optimal for L 2 ranging from 20 to 65 db SPL and for frequencies from 1 to 8 khz. Utilizing a more time-efficient method of varying L 1 continuously, Neely et al. (2005) explored a broad range of L 2 values for frequencies ranging from 1 to 8 khz and suggested the formula L 1 = 0.45L [db SPL]. In addition to recommending greater differences between L 1 and L 2 than those of Kummer et al. (1998), the authors also identified an effect of frequency on optimization formula parameters, grounding it in theory of frequency-selective basilar membrane mechanics (Ruggero et al., 1997; Reichenbach and Hudspeth, 2014). Johnson et al. (2006) suggested that if L DP is maximized via optimized overlap of excitation patterns on the BM, then all parameters affecting the excitation patterns should be varied simultaneously. To that end, they varied both the levels and the frequency ratio of the primary tones over a much wider range of values than had previously been done (Gaskill and Brown, 1990; Abdala, 1996), developing an optimization formula with both level- and frequency-specific components. This result is consistent with the premise that overlap near the f 2 characteristic place is affected by raising or lowering the peak of the excitation pattern at the f 1 place via L 1 modifications, shifting the f 1 place itself via f 2 /f 1 ratio modifications, or some combination of both. Though Johnson et al. (2006) reported that DPOAE levels in response to such a complex stimulation paradigm either matched or exceeded those obtained using the recommendations of either Kummer et al. (1998) or Neely et al. (2005), the effect was generally small and did not appear to provide consistent benefit (Johnson et al., 2010; Kirby et al., 2011). Attempting to control for the effects of the coherent-reflection mechanism, such as through the use of suppressor tones, has likewise led to mixed results (Kirby et al., 2011). 24

43 2.1 Introduction Despite great progress in the theory and practice of evoking DPOAEs, several important factors have yet to be properly accounted for in L 1OP T recommendations. First, optimization formulas do not currently account for the imperfect repeatability of the DPOAE itself. Formulas to predict L 1OP T are traditionally derived via linear regression through the L 1 points found to have evoked the largest DPOAEs for the various L 2. To date, all studies appear to have considered a given L 1 superior to its neighboring values if its associated L DP was higher by as little as 0.1 db. This method neglects variability attributable to the stimulating system, the physiological processes which create the DPOAE, as well as the recording of DPOAEs. Wagner et al. (2008) reported data on the repeatability of DPOAE measurements without probe replacement and with stimuli characteristics held constant, finding a mean standard error of measurement (SEM) across frequencies of 0.67 db. SEM can be interpreted as the standard deviation of the L DP distribution which could be expected if a given test condition were repeated numerous times. These results call into question the practice of accepting minimal differences in L DP in optimization formula creation, as L 1 leading to statistically equivalent L DP would then have to be discarded. Second, determining the presence and influence of frequency effects on optimization formula parameters appears to have traditionally been performed via visual inspection of data sets. As the presence of frequency effects would strengthen the integration of theories of BM mechanics with those of DPOAE generation, it is worthwhile to evaluate results for effects of frequency using appropriate statistics. Furthermore, should such effects be found, it remains unclear if their inclusion, given the real-world limits of stimulation precision achievable with clinical OAE systems, would result in actionable differences in L 1opt when compared with frequency-independent recommendations. This study was conducted in order to: 1) Assess reliability of DPOAEs when 25

44 OPTIMAL DPOAE PRIMARY TONE LEVELS evoked without probe replacement to determine the smallest significant change in L DP. 2) Develop an L 1 optimization formula incorporating findings from objective 1 and evaluate it for effects of frequency. 2.2 Methods Participants Participants were normal hearing adults, with normal hearing defined as airconduction thresholds at or below 15 db HL (IEC 60655, 1979), as measured with ER-3A insert earphones (Etymotic Research, Elk Grove, IL), for audiometric test frequencies between and 8 khz. No participants exhibited a significant air-bone gap (ABG), with ABG defined as a difference between air-conduction and bone-conduction thresholds exceeding 10 db at any octave frequency between 0.5 and 4 khz. Middle ear function was screened via 226-Hz tympanometry using an Interacoustics A/S Titan (Middelfart, Denmark). A given ear passed the screen if it exhibited tympanometric peak pressure between -100 and +50 dapa and peakcompensated static acoustic admittance between 0.3 and 1.5 mmhos (Roup et al., 1998). Otoscopy was completed to confirm that ear canals were free of cerumen. All participants denied a history of middle ear infection, noise exposure, tinnitus, and any other otologic symptoms. Participants were non-randomly assigned to either the first or second experiment based solely upon personal time constraints. Eleven participants (21 ears) between the ages of 20 and 44 years (mean = 24.4 years, SD = 2.8 years) were enrolled in and completed the first experiment. Thirty participants (57 ears) between the ages of 21 and 33 years (mean = 25.5 years, SD = 2.6 years) were enrolled in and completed the second experiment. Research methods for both 26

45 2.2 Methods experiments were approved by the Institutional Review Board of the University Hospital Regensburg Equipment & stimuli The commercially available Echoport ILO292-II otoacoustic emission system with a GD TE+DPOAE probe (Otodynamics, Hatfield, UK) was used for stimulus generation and calibration, DPOAE recording, and response analysis. A point fast-fourier Transform (FFT) was used to analyze responses, resulting in a bin size of approximately 12 Hz. Noise level was defined as the average level in db of the five FFT bins on either side of the bin containing 2f1-f2. Signal level was defined as the level in db of the bin containing 2f 1 -f 2, after the acoustic pressure of the noise had been subtracted from that of the signal. The system was controlled via the Windows 7-based ILOv6 software package installed on a PC. All measurements were performed in a sound-treated booth. Calibration of primary tones was conducted in-situ at the plane of the probe using an SPL-based method and a chirp stimulus. Concerns have been expressed regarding the potential impact of standing waves on the calibrated level of primary tones when using this method (Gilman and Dirks, 1986; Siegel and Hirohata, 1994); however, it was selected for the following reasons. First, participants with a wide range of ear canal dimensions were tested in this study, thereby reducing the systematic impact of standing waves. Second, the resilience against standing waves in the frequency range of interest offered by other calibration methods, such as the Sound Intensity Level (Neely and Gorga, 1998) or Forward Pressure Level methods (Neely and Gorga, 2010), is not yet well-established and potentially small (Burke et al., 2010; Rogers et al., 2010; Kirby et al., 2011; Reuven et al., 2013). Third, the SPL method is utilized by a significant majority of clinical OAE measurement 27

46 OPTIMAL DPOAE PRIMARY TONE LEVELS systems and therefore represents the current standard, if imperfect Evaluation of probe signal stability As reliability of evoked DPOAEs decreases with increases in the variability of primary tone stimuli over the course of a measurement or across measurements, initial technical measurements within a 2 cc coupler were conducted. The DPOAE probe was sealed into one end of a hard-walled, plastic tube with a residual distance from the probe tip to the other end of 20 mm, approximating the separation observed when DPOAEs are measured in adult ears (Siegel and Hirohata, 1994). The microphone of a Bruel-Kjaer 2236 sound level meter (Type 1) was sealed into the tube opposite the probe. For frequencies 1, 2, 3, 4, and 6 khz, stimulus level was varied from 20 to 75 db SPL in 5 db steps, with stimulation lasting 60 seconds at each level. Third-octave filtering around the frequencies of interest and a slow time constant (1 second) were activated. Maximum deviation of the sound pressure level registered by the sound level meter from its value after 3 seconds of stimulation was recorded for each level. The first 3 seconds of stimulation were not included to allow for stabilization of the initial level reading. This process was repeated 3 times per level for each of 2 probe channels, resulting in a total of 180 measurements Experiment 1: Evaluation of L DP reliability In an effort to quantify the magnitude of L DP variability in the absence of stimuli changes, DPOAEs were recorded at f 2 = 1, 2, 3, 4, and 6 khz with f 2 /f 1 = 1.22 and L 1 = L 2 = 65 db SPL in 21 normal hearing ears. Stimulus pairs were presented a minimum of 10 seconds each, with response averaging continuing, if necessary, until a 12 db signal-to-noise (SNR) ratio was achieved. Measurements 28

47 2.2 Methods at each frequency were conducted twice without replacement or re-calibration of the probe. The standard error of measurement was calculated for each frequency and taken as a measure of L DP reliability. SEM = s 1 r, where s is the standard deviation of the combined baseline and follow-up measurements and r represents the correlation between the baseline and follow-up measurements. Accepting the assumption of a normal distribution and requiring a confidence level of 95%, a level difference between two measurements of at least 1.96 SEM is needed before it can be accepted as greater than the effects of test-retest reliability Experiment 2: Development of an optimization formula The second experiment was conducted in order to identify the average L 1 -L 2 differences resulting in maximally evoked DPOAEs for a broad range of frequencies and stimulus levels. The data were collected in 57 normal hearing ears for f 2 = 1, 2, 3, 4, and 6 khz with f 2 /f 1 = 1.22, while L 2 was varied from 20 to 75 db SPL in 5 db steps. For each discrete L 2, L 1 was stimulated according to the formula L 1 = 0.4L [db SPL] (Kummer et al., 1998), as well as up to 15 db above and below this point in 3 db steps. Stimulation 15 db above the recommended level was not possible for values above 65 db SPL due to the probe s upper output limit of 80 db SPL. Measurements within an artificial ear simulator (Type IEC 60711) revealed no significant impact of system distortion for any test frequency at 80 db SPL or below. Stimuli were presented for a minimum of 20 seconds, with response 29

48 OPTIMAL DPOAE PRIMARY TONE LEVELS averaging continuing, if necessary, until a 12 db SNR was achieved or the noise level fell below -20 db SPL. The L 1 of a given primary tone pair was defined as L 1OP T if it fulfilled the following conditions. First, SNR for the measurement of interest, as well as for the measurements immediately preceding and following in the L 2 series, was 12 db. This requirement reduces the potential impact of undetected system distortions on outcomes and is consistent with previous work (Kummer et al., 1998, 2000). A result of this condition, however, is that the maximum L DP for a given L 2 could not be associated with either the highest or lowest L 1 in the L 2 series. While this ensured that only points representing growth function peaks were included in further analyses, it also had the consequence that growth functions exhibiting no saturation within the primary tone level constraints were not included in the analysis. Follow-up testing to assess the impact of excluding these functions revealed no effect on the final optimization formula. Second, the resultant L DP was within 1 db of the maximum L DP obtained from a measurement fulfilling the first condition in the L 2 series. This requirement reduces the bias of only including the L 1 associated with the highest L DP, when test-retest reliability of DPOAE measurements, rather than primary tone level differences, might be responsible for the difference. The criterion of 1 db is in agreement with a previous report of DPOAE reliability without probe replacement (Wagner et al., 2008) Data analysis Consistent with previous work, the primary tone level optimization formula was derived as a linear function (L 1OP T = al 2 + b). For each ear and frequency, L 1OP T were plotted against L 2 and a linear regression analysis was performed to obtain the optimization formula slope and y-intercept. Mean coefficients for each 30